欢迎访问北京化工大学学报(自然科学版),今天是
Email Alert  RSS
管理与数理科学

变时滞反馈控制的混合中立型随机延迟微分方程的指数稳定性

  • 刘琪 ,
  • 兰光强 , *
展开
  • 北京化工大学 数理学院, 北京 100029
兰光强, E-mail:

刘琪, 女, 1998年生, 硕士生

收稿日期: 2022-09-05

  网络出版日期: 2023-12-02

基金资助

北京市自然科学基金(1192013)

版权

版权所有,未经授权。

Exponential stability of hybrid neutral stochastic differential delay equations with time-dependent delay feedback control

  • Qi LIU ,
  • GuangQiang LAN , *
Expand
  • College of Mathematics and Physics, Beijing University of Chemical Technology, Beijing 100029, China

Received date: 2022-09-05

  Online published: 2023-12-02

Copyright

All rights reserved, without authorization

摘要

研究了变时滞反馈控制的混合中立型随机延迟微分方程(HNSDDEs)的指数稳定性。采用函数方法设置合适的变时滞反馈控制函数,得到了该系统的指数稳定性。对比已有的研究成果,本文的主要贡献是在变时滞反馈控制下对HNSDDEs的指数稳定性作了进一步研究。最后,给出一个例子证明了结论的有效性。

本文引用格式

刘琪 , 兰光强 . 变时滞反馈控制的混合中立型随机延迟微分方程的指数稳定性[J]. 北京化工大学学报(自然科学版), 2023 , 50(6) : 105 -111 . DOI: 10.13543/j.bhxbzr.2023.06.013

Abstract

The exponential stability of hybrid neutral stochastic differential delay equations (HNSDDEs) with time-dependent delay feedback control has been investigated. Using the Lyapunov function method, the exponential stability of the system can be obtained by setting an appropriate feedback control function with a variable delay. Compared with the existing research results, the results of this work increase our understanding of the exponential stability of HNSDDEs under the influence of variable delay feedback. Finally, an example is given to prove the validity of the conclusions.

引言

带有变时滞反馈控制的混合中立型随机延迟微分方程(HNSDDEs)常被用于系统未来的建模,目前已经被广泛应用于种群生态、神经网络以及激光器动力学等领域。
对于随机系统突然性的结构变化,常采用连续时间马氏链来描述,带有马氏链的随机延迟微分方程即为混合随机延迟微分方程。文献[1]具体研究了混合随机延迟微分方程,文献[2-4]则进一步考虑了其稳定性及有界性,文献[5-7]又扩展到了带中立项的混合随机延迟微分方程的稳定性研究。
然而并非所有系统都是稳定的,因此设计一个合适的反馈控制使不稳定的系统变得稳定很有意义。相应地,文献[8-11]研究了系统稳定化问题。其中文献[8]研究了常时滞反馈控制的高阶非线性混合随机时滞微分方程的指数稳定性,文献[9]是在文献[10]的基础上进一步研究了变时滞反馈控制的HNSDDEs的Lp渐进稳定性和H稳定性。
本文采用Lyapunov函数方法,进一步研究了变时滞反馈控制下的HNSDDEs的指数稳定性。文献[8]研究了常时滞反馈控制下的混合随机微分延迟方程的指数稳定性,其所涉及的时滞均为常量,本文进一步将常时滞推广到了函数时滞,并且将受控方程推广到了带有中立项的混合随机延迟微分方程,其难点在于找到时滞δ(t)的上界和利用引理2处理中立项。文献[9]研究了变时滞反馈控制的具有时变延迟的高度非线性HNSDDEs的Lp渐近稳定性和H稳定性,但缺少指数稳定性,本文则是通过进一步找到更合适的反馈函数确定了方程的收敛速度,即指数稳定性。

1 基本假设与模型描述

$ \left(\varOmega, \mathscr{F}, \left\{\mathscr{F}_{t}\right\}_{t \geqslant 0}, P\right)$是一个带有σ流(满足通常条件)的完备概率空间, { B(t)}t≥0是定义在其上的m维布朗运动,{r(t)}t≥0是右连马氏链且独立于{ B(t)}t≥0S={1, 2,…,N}是其状态空间,Γ=(γij)N×N是其生成算子。
考虑变时滞反馈控制HNSDDE
$\begin{array}{l} \;\;\;\mathrm{d} \hat{\boldsymbol{x}}(t)=\boldsymbol{f}(\boldsymbol{x}(t), \boldsymbol{x}(t-\tau(t)), t, r(t)) \mathrm{d} t+\boldsymbol{g}(\boldsymbol{x}(t) , \\\boldsymbol{x}(t-\tau(t)), t, r(t)) {\rm{d}}\boldsymbol{B}\left( t \right),t \ge 0\end{array}$
其中$ \hat{\boldsymbol{x}}(t)=\boldsymbol{x}(t)-\boldsymbol{N}(\boldsymbol{x}(t-\boldsymbol{\tau}(t)), t, r(t))$,且初值满足
$\begin{aligned}& \{\boldsymbol{x}(\theta):-\tau \leqslant \theta \leqslant 0\}=\varphi \in C\left([-\tau, 0] ; \mathbb{R}^{n}\right) \\& r(0)=r_{0} \in S\end{aligned}$
其中f, g, N均为Borel可测函数,并且满足
$\begin{array}{l} \boldsymbol{f}: \mathbb{R}^{n} \times \mathbb{R}^{n} \times \mathbb{R}_{+} \times S \rightarrow \mathbb{R}^{n}\\ \boldsymbol{g}: \mathbb{R}^{n} \times \mathbb{R}^{n} \times \mathbb{R}_{+} \times S \rightarrow \mathbb{R}^{n \times m}\\ \boldsymbol{N}: \mathbb{R}^{n} \times \mathbb{R}_{+} \times S \rightarrow \mathbb{R}^{n}\end{array}$
加上反馈控制函数u之后系统变为
$\begin{array}{l} \;\;\;\mathrm{d} \hat{\boldsymbol{x}}(t)=[\boldsymbol{f}(\boldsymbol{x}(t), \boldsymbol{x}(t-\boldsymbol{\tau}(t)), t, r(t))+\boldsymbol{u}(\boldsymbol{x}(t-\\\delta(t)), t, r(t))] \mathrm{d} t+\boldsymbol{g}(\boldsymbol{x}(t), \boldsymbol{x}(t-\tau(t)), t, r(t)) \cdot\\\mathrm{d} \boldsymbol{B}(t), t \geqslant 0\end{array}$
其中0≤δ(t)≤δτ, 0≤τ(t)≤τ。假设
$\begin{array}{l} \boldsymbol{f}(0, 0, t, i)=\boldsymbol{N}(0, t, i) \equiv 0, \boldsymbol{g}(0, 0, t, i) \equiv 0\\V(\boldsymbol{x}, t, i) \in C^{2, 1}\left(\mathbb{R}^{n} \times \mathbb{R}_{+} \times S ; \mathbb{R}_{+}\right)\end{array}$
为方便起见,简记$ \hat{\boldsymbol{x}}=\boldsymbol{x}-\boldsymbol{N}(\boldsymbol{y}, t, i)$
$ V(\boldsymbol{x}, t, i) \in C^{2, 1}\left(\mathbb{R}^{n} \times \mathbb{R}_{+} \times S ; \mathbb{R}_{+}\right)$定义如下算子L
$\begin{array}{l} \;\;\;\mathrm{L} V(\boldsymbol{x}, \boldsymbol{y}, t, i)=V_{t}(\hat{\boldsymbol{x}}, t, i)+V_{x}^{\mathrm{T}}(\hat{\boldsymbol{x}}, t, i) \boldsymbol{f}(\boldsymbol{x}, \boldsymbol{y}, t , \\i)+\frac{1}{2} {trace}\left[\boldsymbol{g}^{\mathrm{T}}(\boldsymbol{x}, \boldsymbol{y}, t, i) V_{x x}(\hat{\boldsymbol{x}}, t, i) \boldsymbol{g}(\boldsymbol{x}, \boldsymbol{y}, t, i)\right]+\\\sum\limits_{j \in s} \gamma_{i j} V(\hat{\boldsymbol{x}}, t, j)\end{array}$
为得到本文主要结论,提出以下假设。
假设1  对任意l>0,存在Kl>0,使得对任意iS, $ t \in \mathbb{R}_{+}$,且$ |\boldsymbol{x}| \vee|\overline{\boldsymbol{x}}| \vee|\boldsymbol{y}| \vee|\overline{\boldsymbol{y}}| \leqslant l$,满足
$\begin{array}{l} \;\;\;|\boldsymbol{f}(\boldsymbol{x}, \boldsymbol{y}, t, i)-\boldsymbol{f}(\overline{\boldsymbol{x}}, \overline{\boldsymbol{y}}, t, i)| \vee | \boldsymbol{g}(\boldsymbol{x}, \boldsymbol{y}, t, i)-\\\boldsymbol{g}(\overline{\boldsymbol{x}}, \overline{\boldsymbol{y}}, t, i) | \leqslant K_{l}(|\boldsymbol{x}-\overline{\boldsymbol{x}}|+|\boldsymbol{y}-\overline{\boldsymbol{y}}|)\end{array}$
假设2  存在K>0,m1>1, m2≥1,使得对∀x, $ \boldsymbol{y} \in \mathbb{R}^{n}, i \in S, t \in \mathbb{R}_{+}$,有
$\begin{aligned}& |\boldsymbol{f}(\boldsymbol{x}, \boldsymbol{y}, t, i)| \leqslant K\left(|\boldsymbol{x}|^{m_{1}}+|\boldsymbol{y}|^{m_{1}}+1\right) \\& |\boldsymbol{g}(\boldsymbol{x}, \boldsymbol{y}, t, i)| \leqslant K\left(|\boldsymbol{x}|^{m_{2}}+|\boldsymbol{y}|^{m_{2}}+1\right)\end{aligned}$
假设3  系统(3)中的时滞函数$ \tau: \mathbb{R}_{+} \rightarrow[0, \tau]$满足
$\tau^{\prime}(t)=\frac{\mathrm{d} \tau(t)}{\mathrm{d} t} \leqslant \bar{\tau}<1, t \geqslant 0$
系统(3)反馈控制函数中的$ \delta: \mathbb{R}_{+} \rightarrow[0, \delta]$满足
$\delta^{\prime}(t)=\frac{\mathrm{d} \delta(t)}{\mathrm{d} t} \leqslant \bar{\delta}<1, t \geqslant 0$
假设4  存在κ∈(0, 1)使得对∀x, $ \boldsymbol{y} \in \mathbb{R}^{n}, i \in S, t \in \mathbb{R}_{+}$,有
$|\boldsymbol{N}(\boldsymbol{x}, t, i)-\boldsymbol{N}(\boldsymbol{y}, t, i)| \leqslant \kappa(1-\bar{\tau})|\boldsymbol{x}-\boldsymbol{y}| $
并且N(0, t, i)≡0。
假设5  存在常数c1, c2, c3, c4>0, c2c3+c4和函数$ V \in C^{2, 1}\left(\mathbb{R}^{n} \times \mathbb{R}_{+} \times S ; \mathbb{R}_{+}\right)$, U1, U2$ C(\mathbb{R} \times \left.[-\tau, +\infty] ; \mathbb{R}_{+}\right)$, 使得对$ \forall \boldsymbol{x}, \boldsymbol{y} \in \mathbb{R}^{n}, i \in S, t \in \mathbb{R}_{+}$,有
$\begin{array}{l} \;\;\;U_{1}(\boldsymbol{x}, t) \leqslant V(\boldsymbol{x}, t, i) \leqslant U_{2}(\boldsymbol{x}, t)\\\;\;\;\mathrm{L}V(\boldsymbol{x}, \boldsymbol{y}, t, i)+V_{x}(\boldsymbol{x}-\boldsymbol{N}(\boldsymbol{y}), t, i) \boldsymbol{u}(\boldsymbol{z}, t, i) \leqslant\\c_{1}-c_{2} U_{2}(\boldsymbol{x}, t)+c_{3}(1-\bar{\tau}) U_{2}(\boldsymbol{y}, t-\tau(t))+c_{4}(1-\\\bar{\delta}) U_{2}(\boldsymbol{z}, t-\delta(t))\end{array} $
由文献[7]可得如下引理。
引理1  设假设1~4成立,且假设5对于$ U_{1}(\boldsymbol{x}, t)=|\boldsymbol{x}|^{\bar{w}}$成立,那么系统(3)有唯一的全局解,并且满足
$\sup \limits_{-\tau \leqslant t<\infty} E|\boldsymbol{x}(t)|^{\bar{w}}<\infty, \bar{w} \geqslant 2\left(m_{1} \vee m_{2}\right)$
由文献[5]中引理2.2以及式(9)可得
引理2  若p≥1,则
$\begin{array}{l} \;\;\;[1-\kappa(1-\bar{\tau})]^{p-1}\left[|\boldsymbol{x}|^{p}-\kappa(1-\overline{{\tau}})|\boldsymbol{y}|^{p}\right] \leqslant\\|\boldsymbol{x}-\boldsymbol{N}(\boldsymbol{y}, t, i)|^{p} \leqslant[1+\kappa(1-\bar{\tau})]^{p-1}\left[|\boldsymbol{x}|^{p}+\kappa(1-\right.\\\left.\bar{\tau})|\boldsymbol{y}|^{p}\right]\end{array}$

2 主要结论与证明

定义片段过程
$\overline{\boldsymbol{x}}(t)=\{\boldsymbol{x}(t+s):-2 \tau \leqslant s \leqslant 0, 0 \leqslant t \leqslant 2 \tau\}$
同理定义${\bar r}$(t),且令
$\left\{\begin{array}{l}r(s)=r(0), s \in[-2 \tau, 0) \\ \boldsymbol{x}(s)=\varphi(-\tau), s \in[-2 \tau, -\tau)\end{array}\right.$
$ U \in C^{2, 1}\left(\mathbb{R}^{n} \times \mathbb{R}_{+} \times S ; \mathbb{R}_{+}\right)$且满足
$\lim \limits_{|\boldsymbol{x}| \rightarrow \infty}\left[\inf \limits_{(t, i) \in \mathbb{R}_{+} \times S} U(\boldsymbol{x}, t, i)\right]=\infty$
对于$ t \in \mathbb{R}_{+}$,定义
$\begin{array}{l} \;\;\;V(\overline{\boldsymbol{x}}(t), t, \bar{r}(t))=U(\hat{\boldsymbol{x}}(t), t, r(t))+\rho \int_{-\delta}^{0} \int_{t+s}^{t} J(v) \cdot\\\mathrm{d} v \mathrm{~d} s\end{array}$
其中ρ>0,且
$\begin{array}{l} \;\;\;J(t):=\delta | \boldsymbol{u}(\boldsymbol{x}(t-\delta(t)), t, r(t))+\boldsymbol{f}(\boldsymbol{x}(t), \boldsymbol{x}(t-\\\tau(t)), t, r(t))\left.\right|^{2}+|\boldsymbol{g}(\boldsymbol{x}(t), \boldsymbol{x}(t-\tau(t)), t, r(t))|^{2}\end{array}$
对于$ \boldsymbol{x}, \boldsymbol{y} \in \mathbb{R}^{n}, i \in S, s \in[-2 \tau, 0)$,设
$\begin{array}{l} \boldsymbol{f}(\boldsymbol{x}, \boldsymbol{y}, s, i) \equiv \boldsymbol{f}(\boldsymbol{x}, \boldsymbol{y}, 0, i)\\\boldsymbol{g}(\boldsymbol{x}, \boldsymbol{y}, s, i) \equiv \boldsymbol{g}(\boldsymbol{x}, \boldsymbol{y}, 0, i)\\\boldsymbol{u}(\boldsymbol{z}, s, i) \equiv \boldsymbol{u}(\boldsymbol{z}, 0, i)\end{array}$
由伊藤公式可得
$\begin{array}{l} \;\;\; \mathrm{d} U(\hat{\boldsymbol{x}}(t), t, r(t))=\left[U_{t}(\hat{\boldsymbol{x}}(t), t, r(t))+\right.\\U_{x}^{\mathrm{T}}(\hat{\boldsymbol{x}}(t), t, r(t))(f(\boldsymbol{x}(t), \boldsymbol{x}(t-\tau(t)), t, r(t))+\\\boldsymbol{u}(\boldsymbol{x}(t-\delta(t)), t, r(t)))+\sum\limits_{j \in S} \gamma_{j, r(t)} U(\hat{\boldsymbol{x}}(t), t, j)+\\\frac{1}{2} {trace}\left[\boldsymbol{g}^{\mathrm{T}}(\boldsymbol{x}(t), \boldsymbol{x}(t-\tau(t)), t, r(t)) U_{x x}(\hat{\boldsymbol{x}}(t), t\right. , \\r(t)) \boldsymbol{g}(\boldsymbol{x}(t), \boldsymbol{x}(t-\tau(t)), t, r(t))] \mathrm{d} t+\mathrm{d} \boldsymbol{B}(t) \end{array}$
其中,B(t)是局部鞅,并且B(0)=0。整理式(13)得
$\begin{array}{l} \;\;\;\mathrm{d} U(\hat{\boldsymbol{x}}(t), t, r(t))=\ell U(\boldsymbol{x}(t), \boldsymbol{x}(t-\tau(t)), t, \\r(t)) \mathrm{d} t+U_{x}^{\mathrm{T}}(\hat{\boldsymbol{x}}(t), t, r(t))[\boldsymbol{u}(\boldsymbol{x}(t-\delta(t)), t , \\r(t))-\boldsymbol{u}(\boldsymbol{x}(t), t, r(t))] \mathrm{d} t+\mathrm{d} \boldsymbol{B}(t)\end{array}$
其中,
$\begin{array}{l} \;\;\;\ell U(\boldsymbol{x}(t), \boldsymbol{x}(t-\tau(t)), t, r(t))=U_{t}(\hat{\boldsymbol{x}}(t), t , \\r(t))+U_{x}^{\mathrm{T}}(\hat{\boldsymbol{x}}(t), t, r(t))[\boldsymbol{f}(\boldsymbol{x}(t), \boldsymbol{x}(t-\tau(t)), t, \\r(t))+\boldsymbol{u}(\boldsymbol{x}(t), t, r(t))]+\sum\limits_{j \in S} \gamma_{j, r(t)} U(\hat{\boldsymbol{x}}(t), t, j)+\\\frac{1}{2} {trace}\left[\boldsymbol{g}^{\mathrm{T}}(\boldsymbol{x}(t), \boldsymbol{x}(t-\tau(t)), t, r(t)) U_{x x}(\hat{\boldsymbol{x}}(t), t\right., \\r(t)) \boldsymbol{g}(\boldsymbol{x}(t), \boldsymbol{x}(t-\tau(t)), t, r(t))]\end{array}$
进而易得以下结论。
引理3  $ V(\overline{\boldsymbol{x}}(t), t, \bar{r}(t)), t \geqslant 0$是伊藤过程,且有
$\begin{array}{l} \;\;\;\mathrm{d} V(\overline{\boldsymbol{x}}(t), t, \bar{r}(t))=\mathrm{d} \boldsymbol{B}(t)+\mathrm{L} V(\overline{\boldsymbol{x}}(t), t, \bar{r}(t)) \cdot\\\mathrm{d} t\end{array}$
其中,
$\begin{array}{l} \;\;\;\mathrm{L}V(\overline{\boldsymbol{x}}(t), t, \bar{r}(t))=\ell U(\boldsymbol{x}(t), \boldsymbol{x}(t-\tau(t)), t , \\r(t))+\rho \delta J(t)-\rho \int_{t-\delta}^{t} J(v) \mathrm{d} v+U_{x}^{\mathrm{T}}(\hat{\boldsymbol{x}}(t), t, r(t)) \cdot\\[\boldsymbol{u}(\boldsymbol{x}(t-\delta(t)), t, r(t))-\boldsymbol{u}(\boldsymbol{x}(t), t, r(t))]\end{array}$
假设6  对于函数$ \boldsymbol{u}: \mathbb{R}^{n} \times S \times \mathbb{R}_{+} \rightarrow \mathbb{R}^{n}$,存在实数ai, ${\bar a}$i,正数di, ${\bar d}$i和非负数bi, ${\bar b}$i, ei, ${\bar e}$i(iS),对于任意q1>1, p>2有
$\begin{array}{l} \;\;\;\boldsymbol{x}^{\mathrm{T}}[\boldsymbol{f}(\boldsymbol{x}, \boldsymbol{y}, t, i)+\boldsymbol{u}(\boldsymbol{x}, t, i)]+\frac{1}{2}|\boldsymbol{g}(\boldsymbol{x}, \boldsymbol{y}, t, i)|^{2} \leqslant\\a_{i}|\boldsymbol{x}|^{2}+b_{i}|\boldsymbol{y}|^{2}-d_{i}|\boldsymbol{x}|^{p}+e_{i}|\boldsymbol{y}|^{p}\\\;\;\;\boldsymbol{x}^{\mathrm{T}}[\boldsymbol{f}(\boldsymbol{x}, \boldsymbol{y}, t, i)+\boldsymbol{u}(\boldsymbol{x}, t, i)]+\frac{q_{1}}{2}|\boldsymbol{g}(\boldsymbol{x}, \boldsymbol{y}, t, i)|^{2} \leqslant\\\bar{a}_{i}|\boldsymbol{x}|^{2}+\bar{b}_{i}|\boldsymbol{y}|^{2}-\bar{d}_{i}|\boldsymbol{x}|^{p}+\bar{e}_{i}|\boldsymbol{y}|^{p}\end{array}$
$\begin{aligned}& \boldsymbol{A}_{1}:=-2 {diag}\left(a_{1}, a_{2}, \cdots, a_{N}\right)-\varGamma \\& \boldsymbol{A}_{2}:=-\left(q_{1}+1\right) {diag}\left(\bar{a}_{1}, \bar{a}_{2}, \cdots, \bar{a}_{N}\right)-\varGamma\end{aligned}$
是非奇异M矩阵(具体定义可参考文献[1]中的2.6部分),并有
$\begin{aligned}& 1>\gamma_{1}, \gamma_{2}>\gamma_{3}, 1>\gamma_{4}, \gamma_{5}>\gamma_{6} \\& \left(\theta_{1}, \theta_{2}, \cdots, \theta_{N}\right)^{\mathrm{T}}=\boldsymbol{A}_{1}^{-1}(1, \cdots, 1)^{\mathrm{T}} \\& \left(\bar{\theta}_{1}, \bar{\theta}_{2}, \cdots, \bar{\theta}_{N}\right)^{\mathrm{T}}=\boldsymbol{A}_{2}^{-1}(1, \cdots, 1)^{\mathrm{T}} \\& \gamma_{1}=\max _{i \in S} 2 \theta_{i} b_{i}, \gamma_{2}=\min _{i \in S} 2 \theta_{i} d_{i} \\& \gamma_{3}=\max _{i \in S} 2 \theta_{i} e_{i}, \gamma_{4}=\max _{i \in S}\left(q_{1}+1\right) \bar{\theta}_{i} \bar{b}_{i}\\& \gamma_{5}=\min _{i \in S}\left(q_{1}+1\right) \bar{\theta}_{i} \bar{d}_{i}, \gamma_{6}=\max _{i \in S}\left(q_{1}+1\right) \bar{\theta}_{i} \bar{e}_{i}\end{aligned}$
其中θiθi是正数。
需要注意的是,关于控制函数u的选取,考虑如下特殊情况
$\begin{array}{l} \;\;\;\boldsymbol{x}^{\mathrm{T}} \boldsymbol{f}(\boldsymbol{x}, \boldsymbol{y}, t, i)+\frac{q-1}{2}|\boldsymbol{g}(\boldsymbol{x}, \boldsymbol{y}, t, i)|^{2} \leqslant a\left(|\boldsymbol{x}|^{2}+\right.\\\left.|\boldsymbol{y}|^{2}\right)-b|\boldsymbol{x}|^{p}+c|\boldsymbol{y}|^{p}\end{array}$
其中a>0, bc>0。由于|x|2, |y|2的系数均为正数,因此只能得到原方程的矩有界性,而得不到稳定性。此时可选取u(x, t, i)= Ax,其中矩阵A为实对称正定矩阵,且满足λmax (A)<-2a,从而
$\begin{array}{l} \;\;\;\boldsymbol{x}^{\mathrm{T}}[\boldsymbol{f}(\boldsymbol{x}, \boldsymbol{y}, t, i)+\boldsymbol{u}(\boldsymbol{x}, t, i)]+\frac{q-1}{2} \cdot\\|\boldsymbol{g}(\boldsymbol{x}, \boldsymbol{y}, t, i)|^{2} \leqslant\left(\lambda_{\max }(\boldsymbol{A})+a\right)|\boldsymbol{x}|^{2}+a|\boldsymbol{y}|^{2}-\\b|\boldsymbol{x}|^{p}+c|\boldsymbol{y}|^{p}\end{array}$
故加上控制项之后的系统指数稳定。
假设7  存在$ U \in C^{2, 1}\left(\mathbb{R}^{n} \times \mathbb{R}_{+} \times S ; \mathbb{R}_{+}\right), H \in$$ C\left(\mathbb{R}^{n} ; \mathbb{R}_{+}\right)$, 及常数0<α<1, 0<βλ, 0<λ1, λ2, λ3, ρ1, ρ2,使得对任意的$ \boldsymbol{x}, \boldsymbol{y} \in \mathbb{R}^{n}, i \in S, t \in \mathbb{R}_{+}$
$\begin{array}{l}\;\;\; \ell U(\boldsymbol{x}, \boldsymbol{y}, t, i)+\lambda_{1}\left|U_{x}(\hat{\boldsymbol{x}}, t, i)\right|^{2}+\lambda_{2} \cdot\\|\boldsymbol{f}(\boldsymbol{x}, \boldsymbol{y}, t, i)|^{2}+\lambda_{3}|\boldsymbol{g}(\boldsymbol{x}, \boldsymbol{y}, t, i)|^{2} \leqslant-\lambda|\boldsymbol{x}|^{2}+(1-\\\bar{\tau}) \beta|\boldsymbol{y}|^{2}-H(\boldsymbol{x})+(1-\overline{\boldsymbol{\tau}}) \alpha H(\boldsymbol{y})\end{array}$
其中,ρ1|x|p+q1-1H(x)≤ρ2(1+|x|p+q1-1)。
假设8  存在λ4>0满足
$|\boldsymbol{u}(\boldsymbol{x}, t, i)-\boldsymbol{u}(\boldsymbol{y}, t, i)| \leqslant \lambda_{4}|\boldsymbol{x}-\boldsymbol{y}|$
并且有u(0, t, i)=0。故有$ \forall \boldsymbol{x} \in \mathbb{R}^{n}, \boldsymbol{u}(\boldsymbol{x}, t, i) \leqslant \lambda_{4} \cdot |\boldsymbol{x}|$
定理1  令${\bar q}$ ∈[2, ${\bar w}$), ${\bar w}$≥2(m1m2)。若假设1~8成立,且常数满足
$\begin{array}{l} \;\;\;\;\kappa(1-\bar{\tau})<\sqrt{\frac{1}{2}} \\ \;\;\;\;\delta \leqslant \frac{\sqrt{\lambda_{1} \lambda_{2}(1-\kappa)(1-\kappa(1-\bar{\tau}))}}{\lambda_{4}} \wedge \\\frac{2 \lambda_{1} \lambda_{3}(1-\kappa)(1-\kappa(1-\bar{\tau}))}{\lambda_{4}^{2}} \wedge\\\frac{\sqrt{(\lambda-\beta)(1-\delta) \lambda_{1}(1-\kappa)(1-\kappa(1-\bar{\tau}))}}{\lambda_{4}^{2}}\end{array}$
则对任意初值,存在ε>0使得系统(3)的解满足
$\mathop {\lim }\limits_{t \to \infty } \sup \frac{1}{t} \ln \left(E|\boldsymbol{x}(t)|^{\bar{q}}\right) \leqslant-\varepsilon \frac{\bar{w}-\bar{q}}{\bar{w}-2}$
其中ε=ε1ε2ε3ε4ε1, ε2, ε3, ε4分别是以下4个方程的根
$\begin{array}{l} \;\;\;\varepsilon \delta+2(1-\kappa)(1-\kappa(1-\bar{\tau}))=1\\\;\;\;\left[\varepsilon h_{3} \rho_{1}^{-1}(1+\kappa(1-\bar{\tau}))^{p+q_{1}-2}\right]\left(\kappa \mathrm{e}^{\varepsilon \tau}+1\right)+\mathrm{e}^{\varepsilon \tau} \alpha=1\\\;\;\;\varepsilon\left(h_{2}+h_{3}\right)(1+\kappa(1-\bar{\tau}))\left(1+\mathrm{e}^{\varepsilon \tau} \kappa\right)+\beta \mathrm{e}^{\varepsilon \tau}+\\\frac{2 \rho \delta^{2} \lambda_{4}^{2} \mathrm{e}^{\varepsilon \delta}}{1-\bar{\delta}}+\frac{\lambda_{4} \kappa^{2}(1-\bar{\tau}) \mathrm{e}^{\varepsilon \tau}\left(1-\bar{\tau}-\bar{\delta}+\mathrm{e}^{\varepsilon \delta}(1-\bar{\tau})\right)}{\lambda_{1}(1-\bar{\delta}-\bar{\tau})}=\\\lambda\\\;\;\;\;2 \mathrm{e}^{\varepsilon \tau} \kappa^{2}(1-\bar{\tau})^{2}=1\end{array}$
特别地,当${\bar q}$=2时有
$\mathop {\lim }\limits_{t \to \infty } \sup \frac{1}{t} \ln \left(E|\boldsymbol{x}(t)|^{2}\right) \leqslant-\varepsilon$
即满足均方指数稳定。
证明:证明分为两步。
1) 第一步
k0>0足够大使得$ \|\varphi\|:=\sup \limits_{-\tau \leqslant s \leqslant 0} \varphi(s)<k_{0}$。定义σk=inf {t≥0:| x(t)≥k|}(kk0),且inf ϕ=∞。由引理1和文献[7],当k→∞,则σk→∞, a.s.根据假设6再定义
$U(\hat{\boldsymbol{x}}, i)=\theta_{i}|\hat{\boldsymbol{x}}|^{2}+\bar{\theta}_{i}|\hat{\boldsymbol{x}}|^{q_{1}+1}$
由伊藤公式有
$\begin{array}{l} \;\;\;\mathrm{e}^{\varepsilon t} E V(\overline{\boldsymbol{x}}(t), t, \bar{r}(t))=V(\overline{\boldsymbol{x}}(0), 0, \bar{r}(0))+\\\int_{0}^{t} \mathrm{e}^{\varepsilon s}(\varepsilon V(\overline{\boldsymbol{x}}(s), s, \bar{r}(s))+\mathrm{L} V(\overline{\boldsymbol{x}}(s), s, \bar{r}(s))) \mathrm{d} s\end{array}$
$ h_{1}=\min \limits_{i \in S} \theta_{i}, h_{2}=\max \limits_{i \in S} \theta_{i}, h_{3}=\max \limits_{i \in S} \bar{\theta}_{i}$,结合式(12)可得
$\begin{array}{l} \;\;\;h_{1} \mathrm{e}^{\varepsilon\left(t \wedge \sigma_{k}\right)} E\left|\hat{\boldsymbol{x}}\left(t \wedge \sigma_{k}\right)\right|^{2} \leqslant V(\overline{\boldsymbol{x}}(0), 0, \bar{r}(0))+\\\int_{0}^{t \wedge \sigma_{k}} \mathrm{e}^{\varepsilon s} E(\mathrm{L}V(\overline{\boldsymbol{x}}(s), s, \bar{r}(s))) \mathrm{d} s+\varepsilon \rho J_{1}\left(t \wedge \sigma_{k}\right)+\\\int_{0}^{t \wedge \sigma_{k}} \mathrm{e}^{\varepsilon s}\left(\varepsilon h_{2} E|\hat{\boldsymbol{x}}(s)|^{2}+\varepsilon h_{3} E|\hat{\boldsymbol{x}}(s)|^{q_{1}+1}\right) \mathrm{d} s\end{array}$
其中,$ J_{1}\left(t \wedge \sigma_{k}\right)=E \int_{0}^{t \wedge \sigma_{k}} \mathrm{e}^{\varepsilon s}\left(\int_{-\delta}^{0} \int_{s+u}^{s} J(v) \mathrm{d} v \mathrm{~d} u\right) \cdot \mathrm{d} s$
对于式(20)中的$ E\left|\hat{\boldsymbol{x}}\left(t \wedge \sigma_{k}\right)\right|^{2}$结合基本不等式可得到
$\begin{array}{l} \;\;\;E\left|\boldsymbol{x}\left(t \wedge \sigma_{k}\right)\right|^{2} \leqslant 2 E\left|\hat{\boldsymbol{x}}\left(t \wedge \sigma_{k}\right)\right|^{2}+2 \kappa^{2}(1-\\\bar{\tau})^{2} E\left|\boldsymbol{x}\left(t \wedge \sigma_{k}-\tau\left(t \wedge \sigma_{k}\right)\right)\right|^{2}\end{array}$
对于式(20)中的$ \mathrm{L} V(\overline{\boldsymbol{x}}(t), t, \bar{r}(t))$结合式(14)和假设7有
$\begin{array}{l} \;\;\;\mathrm{L} V(\overline{\boldsymbol{x}}(t), t, \bar{r}(t)) \leqslant-\lambda|\boldsymbol{x}(t)|^{2}+(1-\bar{\tau}) \beta \cdot \\|\boldsymbol{x}(t-\tau(t))|^{2}-H(\boldsymbol{x}(t))+(1-\bar{\tau}) \alpha H(\boldsymbol{x}(t-\\\tau(t)))-\lambda_{1}\left|U_{x}(\hat{\boldsymbol{x}}(t), t, r(t))\right|^{2}-\lambda_{2} | \boldsymbol{f}(\boldsymbol{x}(t), \\\boldsymbol{x}(t-\tau(t)), t, r(t))\left.\right|^{2}-\lambda_{3} | \boldsymbol{g}(\boldsymbol{x}(t), \boldsymbol{x}(t-\tau(t)), t, \\r(t))\left.\right|^{2}+\rho \delta J(t)-\rho \int_{t-\delta}^{t} J(v) \mathrm{d} v+U_{x}^{\mathrm{T}}(\hat{\boldsymbol{x}}(t), t, r(t)) \cdot \\[\boldsymbol{u}(\boldsymbol{x}(t-\delta(t)), t, r(t))-\boldsymbol{u}(\boldsymbol{x}(t), t, r(t))]\end{array}$
由假设8运用均值不等式可以得到
$\begin{array}{l} \;\;\;U_{x}^{\mathrm{T}}(\hat{\boldsymbol{x}}(t), t, r(t))[\boldsymbol{u}(\boldsymbol{x}(t-\delta(t)), t, r(t))-\\\boldsymbol{u}(\boldsymbol{x}(t), t, r(t))] \leqslant \lambda_{1}\left|U_{x}(\hat{\boldsymbol{x}}(t), t, r(t))\right|^{2}+\frac{\lambda_{4}^{2}}{4 \lambda_{1}} \cdot\\|\boldsymbol{x}(t-\delta(t))-\boldsymbol{x}(t)|^{2}\end{array}$
定义$ \rho=\frac{\lambda_{4}^{2}}{2 \lambda_{1}(1-\kappa)(1-\kappa(1-\bar{\tau}))}$,由定理1中δ满足的不等式知2ρδ2λ2, ρδλ3。再由Hölder不等式有
$\begin{array}{l} \;\;\;E|\boldsymbol{x}(t-\delta(t))-\boldsymbol{x}(t)|^{2} \leqslant 2 E | \hat{\boldsymbol{x}}(t)-\hat{\boldsymbol{x}}(t-\\\delta(t))\left.\right|^{2}+2 E | \boldsymbol{N}(\boldsymbol{x}(t-\tau(t)), t, r(t))-\boldsymbol{N}(\boldsymbol{x}(t-\\\tau(t)-\delta(t), t, r(t))\left.\right|^{2} \leqslant 4 E \int_{t-\delta}^{t}[\delta | \boldsymbol{u}(\boldsymbol{x}(v-\delta(v)), v, \\r(v))+\left.\boldsymbol{f}(\boldsymbol{x}(v), \boldsymbol{x}(v-\boldsymbol{\tau}(v)), v, r(v))\right|^{2}+| \boldsymbol{g}(\boldsymbol{x}(v), \\\left.\boldsymbol{x}(v-\tau(v)), v, r(v))\left.\right|^{2}\right] \mathrm{d} v+2 \kappa^{2}(1-\bar{\tau})^{2} E |\boldsymbol{x}(t-\\\tau \left( t \right)) - x\left( {t - \tau \left( t \right) - \delta \left( t \right)} \right){|^2}\end{array} $
所以有
$\begin{array}{l} \;\;\;E \operatorname{L}V(\overline{\boldsymbol{x}}(t), t, \bar{r}(t)) \leqslant-\lambda E|\boldsymbol{x}(t)|^{2}+(1-\overline{\boldsymbol{\tau}}) \cdot\\\beta E|\boldsymbol{x}(t-\tau(t))|^{2}-E H(\boldsymbol{x}(t))+(1-\bar{\tau}) \alpha E H(\boldsymbol{x}(t-\\\tau(t)))+2 \rho \delta^{2} \lambda_{4}^{2} E|\boldsymbol{x}(t-\delta(t))|^{2}+\left(\frac{\lambda_{4}^{2}}{\lambda_{1}}-\rho\right) \cdot\\E \int_{t-\delta}^{t} J(v) \mathrm{d} v+\frac{\lambda_{4} \kappa^{2}(1-\bar{\tau})^{2}}{2 \lambda_{1}} E | \boldsymbol{x}(t-\tau(t))-\boldsymbol{x}(t-\\\tau(t)-\delta(t))\left.\right|^{2}\end{array}$
对于式(20)中的$ E|\hat{\boldsymbol{x}}(t)|^{q_{1}+1}$有以下关系式
$E|\hat{\boldsymbol{x}}(t)|^{q_{1}+1} \leqslant E|\hat{\boldsymbol{x}}(t)|^{2}+E|\hat{\boldsymbol{x}}(t)|^{p+q_{1}-1}$
又由假设7有
$|\boldsymbol{x}(t)|^{p+q_{1}-1} \leqslant \rho_{1}^{-1} H(\boldsymbol{x}(t))$
所以结合式(20)~(23)有
$\begin{array}{l} \;\;\;\frac{1}{2} h_{1} \mathrm{e}^{\varepsilon\left(t \wedge \sigma_{k}\right)} E\left|\boldsymbol{x}\left(t \wedge \sigma_{k}\right)\right|^{2} \leqslant \varPi_{1}+\varPi_{2}+\varPi_{3}+\\\int_{0}^{t \wedge \sigma_{k}} \mathrm{e}^{\varepsilon s}\left(\varepsilon h_{2} E|\hat{\boldsymbol{x}}(s)|^{2}+\varepsilon h_{3} E|\hat{\boldsymbol{x}}(s)|^{2}+\varepsilon h_{3}\right.\cdot \\\left.E|\hat{\boldsymbol{x}}(s)|^{p+q_{1}-1}\right) \mathrm{d} s+\int_{0}^{t \wedge \sigma_{k}} \mathrm{e}^{\varepsilon s} E\left[-\lambda|\boldsymbol{x}(s)|^{2}+(1-\overline{\boldsymbol{\tau}}) \cdot\right.\\\beta|\boldsymbol{x}(s-\tau(s))|^{2}-H(\boldsymbol{x}(s))+(1-\bar{\tau}) \alpha H(\boldsymbol{x}(s-\\\tau(s)))+2 \rho \delta^{2} \lambda_{4}^{2}|\boldsymbol{x}(s-\delta(s))|^{2}+\frac{\lambda_{4} \kappa^{2}(1-\bar{\tau})^{2}}{2 \lambda_{1}}\cdot \\\left.|\boldsymbol{x}(s-\tau(s))-\boldsymbol{x}(s-\boldsymbol{\tau}(s)-\delta(s))|^{2}\right] \mathrm{d} s\end{array}$
其中,
$\begin{array}{l} \;\;\;\varPi_{1}=h_{1} \mathrm{e}^{\varepsilon\left(t \wedge \sigma_{k}\right)} \kappa^{2}(1-\bar{\tau})^{2} E | \boldsymbol{x}\left(t \wedge \sigma_{k}-\tau(t \wedge\right.\\\left.\left.\sigma_{k}\right)\right)\left.\right|^{2}\\\;\;\;\varPi_{2}=V(\overline{\boldsymbol{x}}(0), 0, \bar{r}(0))\\\;\;\;\varPi_{3}=\varepsilon \rho J_{1}\left(t \wedge \sigma_{k}\right)+\left(\frac{\lambda_{4}^{2}}{\lambda_{1}}-\rho\right) J_{2}\left(t \wedge \sigma_{k}\right)\\\;\;\;J_{2}\left(t \wedge \sigma_{k}\right)=E \int_{0}^{t \wedge \sigma_{k}} \mathrm{e}^{\varepsilon s}\left[\int_{s-\delta}^{s} J(v) \mathrm{d} v\right] \mathrm{d} s\end{array}$
易得J1(tσk)≤δJ2(tσk)。取ε1ε1ρδ+$ \frac{\lambda_{4}^{2}}{\lambda_{1}}-\rho=0$的唯一解,则由ρ的定义知,对任意0<εε1,有Π3≤0。结合式(11),令k→∞,结合式(24), 式(25)化为
$\frac{1}{2} h_{1} \mathrm{e}^{\varepsilon t} E|\boldsymbol{x}(t)|^{2} \leqslant \bar{\varPi}_{1}+\varPi_{2}+\varPi_{4}+\varPi_{5}$
其中,
$\begin{array}{l} \;\;\; \bar{\varPi}_{1}=h_{1} \mathrm{e}^{\varepsilon t} \kappa^{2}(1-\bar{\tau})^{2} E|\boldsymbol{x}(t-\tau(t))|^{2} \\\;\;\;\varPi_{4}=\int_{0}^{t} \mathrm{e}^{\varepsilon s}\left\{\varepsilon h_{3} \rho_{1}^{-1}[1+\kappa(1-\bar{\tau})]^{p+q_{1}-2} \right. \cdot \\ {[E H(\boldsymbol{x}(s))+\kappa(1-\bar{\tau}) E H(\boldsymbol{x}(s-\tau(s)))]-} \\ E H(\boldsymbol{x}(s))+(1-\bar{\tau}) \alpha E H(\boldsymbol{x}(s-\tau(s)))\} \mathrm{d} s\\ \;\;\;\varPi_{5}=\int_{0}^{t} \mathrm{e}^{\varepsilon s}\left\{\varepsilon\left(h_{2}+h_{3}\right)[1+\kappa(1-\bar{\tau})]\right. \cdot \\\left.\left[E|\boldsymbol{x}(s)|^{2}+\kappa(1-\bar{\tau}) E|\boldsymbol{x}(s-\tau(s))|^{2}\right]\right\} \mathrm{d} s+\\\int_{0}^{t} \mathrm{e}^{\varepsilon s}\left[-\lambda E|\boldsymbol{x}(s)|^{2}+(1-\bar{\tau}) \beta E|\boldsymbol{x}(s-\tau(s))|^{2}+\right.\\2 \rho \delta^{2} \lambda_{4}^{2} E|\boldsymbol{x}(s-\delta(s))|^{2}+\frac{\lambda_{4} \kappa^{2}(1-\bar{\tau})^{2}}{2 \lambda_{1}} E | \boldsymbol{x}(s-\\\left.\tau(s))-\left.\boldsymbol{x}(s-\tau(s)-\delta(s))\right|^{2}\right] \mathrm{d} s\end{array}$
对于Π2,由初值条件、假设2、假设8、引理2和式(12)得$ V(\overline{\boldsymbol{x}}(0), 0, \bar{r}(0))<\infty$,并且记为C0, C0为常数。
对于Π4,根据假设3化简有
$\begin{array}{l} \;\;\;\varPi_{4} \leqslant\left\{\left[\varepsilon h_{3}(1+\kappa(1-\bar{\tau}))^{p+q_{1}-2} \rho_{1}^{-1}\right]\left(\kappa \mathrm{e}^{\varepsilon \tau}+1\right)+\right.\\\left.\mathrm{e}^{\varepsilon \tau} \alpha-1\right\} \int_{0}^{t} \mathrm{e}^{\varepsilon s} E[H(\boldsymbol{x}(s))] \mathrm{d} s+\mathrm{e}^{\varepsilon \tau}\left[\varepsilon h_{3}(1+\kappa(1-\right.\\\left.\bar{\tau}))^{p+q_{1}-2} \rho_{1}^{-1} \kappa+\alpha\right] \int_{-\tau}^{0} \mathrm{e}^{\varepsilon s} E[H(\boldsymbol{x}(s))] \mathrm{d} s\end{array}$
ε2为[ε2h3(1+κ(1-τ))p+q1-2ρ1-1](κeε2τ+ 1)+eε2τα-1=0的唯一解,则对任意0<εε2以及0<α<1即可满足
$\begin{array}{l} \;\;\;\varPi_{4} \leqslant \mathrm{e}^{\varepsilon \tau}\left[\varepsilon h_{3}(1+\kappa(1-\bar{\tau}))^{p+q_{1}-2} \rho_{1}^{-1} \kappa+\alpha\right] \cdot\\\int_{-\tau}^{0} \mathrm{e}^{\varepsilon s} E[H(\boldsymbol{x}(s))] \mathrm{d} s<\infty\end{array}$
对于Π5,令ε3
$\begin{array}{l} \;\;\;\varepsilon_{3}\left(h_{2}+h_{3}\right)(1+\kappa(1-\bar{\tau}))\left(1+\mathrm{e}^{\varepsilon_{3} \tau} \kappa\right)+\beta \mathrm{e}^{\varepsilon_{3} \tau}+\\\frac{2 \rho \delta^{2} \lambda_{4}^{2} \mathrm{e}^{\varepsilon_{3} \delta}}{1-\bar{\delta}}+\frac{\lambda_{4} \kappa^{2}(1-\bar{\tau}) \mathrm{e}^{\varepsilon_{3} \tau}\left(1-\bar{\tau}-\bar{\delta}+\mathrm{e}^{\varepsilon_{3} \delta}(1-\bar{\tau})\right)}{\lambda_{1}(1-\bar{\delta}-\bar{\tau})}=\\\lambda\end{array}$
的唯一解,对任意0<εε3,有
$\begin{array}{l}\;\;\;\varPi_{5} \leqslant \mathrm{e}^{\varepsilon \tau}\left[\varepsilon\left(h_{2}+h_{3}\right)(1+\kappa(1-\bar{\tau})) \kappa+\beta+\right.\\\left.\frac{\lambda_{4} \kappa^{2}(1-\bar{\tau})}{\lambda_{1}}\right] \int_{-\tau}^{0} \mathrm{e}^{\varepsilon s} E|\boldsymbol{x}(s)|^{2} \mathrm{~d} s+\frac{2 \rho \delta^{2} \lambda_{4}^{2} \mathrm{e}^{\varepsilon \delta}}{1-\bar{\delta}} \int_{-\delta}^{0} \mathrm{e}^{\varepsilon s} \cdot\\E|\boldsymbol{x}(s)|^{2} \mathrm{~d} s+\frac{\lambda_{4} \kappa^{2}(1-\bar{\tau})^{2} \mathrm{e}^{\varepsilon(\tau+\delta)}}{\lambda_{1}(1-\bar{\delta}-\bar{\tau})} \int_{-\delta-\tau}^{0} \mathrm{e}^{\varepsilon s} E|\boldsymbol{x}(s)|^{2} \mathrm{~d} s+\\\left[\varepsilon\left(h_{2}+h_{3}\right)(1+\kappa-\kappa \bar{\tau})\left(1+\mathrm{e}^{\varepsilon \tau} \kappa\right)+\beta \mathrm{e}^{\varepsilon \tau}+\right.\\\frac{2 \rho \delta^{2} \lambda_{4}^{2} e^{\varepsilon \delta}}{1-\bar{\delta}}+\frac{\lambda_{4} \kappa^{2}(1-\bar{\tau}) \mathrm{e}^{\varepsilon \tau}\left(1-\bar{\tau}-\bar{\delta}+\mathrm{e}^{\varepsilon \delta}(1-\bar{\tau})\right)}{\lambda_{1}(1-\bar{\delta}-\bar{\tau})}-\\\lambda] \int_{0}^{t} \mathrm{e}^{\varepsilon s} E|\boldsymbol{x}(s)|^{2} \mathrm{~d} s \leqslant \mathrm{e}^{\varepsilon \tau}\left[\boldsymbol{\varepsilon}\left(h_{2}+h_{3}\right)(1+\kappa(1-\right.\\\left.\bar{\tau})) \kappa+\beta+\frac{\lambda_{4} \kappa^{2}(1-\bar{\tau})}{\lambda_{1}}\right] \int_{-\tau}^{0} \mathrm{e}^{\varepsilon s} E|\boldsymbol{x}(s)|^{2} \mathrm{~d} s+\\\frac{2 \rho \delta^{2} \lambda_{4}^{2} \mathrm{e}^{\varepsilon \delta}}{1-\bar{\delta}} \int_{-\delta}^{0} \mathrm{e}^{\varepsilon s} E|\boldsymbol{x}(s)|^{2} \mathrm{~d} s+\frac{\lambda_{4} \kappa^{2}(1-\bar{\tau})^{2} \mathrm{e}^{\varepsilon(\tau+\delta)}}{\lambda_{1}(1-\bar{\delta}-\bar{\tau})} \cdot\\\int_{-\delta-\tau}^{0} \mathrm{e}^{\varepsilon s} E|\boldsymbol{x}(s)|^{2} \mathrm{~d} s<\infty \end{array}$
综上对任意0<εε1ε2ε3,可得
$\begin{array}{l} \;\;\;\frac{1}{2} h_{1} \mathrm{e}^{\varepsilon t} E|\boldsymbol{x}(t)|^{2} \leqslant h_{1} \mathrm{e}^{\varepsilon t} \kappa^{2}(1-\bar{\tau})^{2} E | \boldsymbol{x}(t-\\\tau(t))\left.\right|^{2}+C_{1}\end{array}$
其中C1是一个常数。
2) 第二步
式(29)经过整理可以得到
$\begin{array}{l} \;\;\; \mathrm{e}^{\varepsilon t} E|\boldsymbol{x}(t)|^{2} \leqslant 2 \mathrm{e}^{\varepsilon \tau} \mathrm{e}^{\varepsilon(t-\tau(t))} \kappa^{2}(1-\overline{\boldsymbol{\tau}})^{2} E | \boldsymbol{x}(t-\\ \tau(t))\left.\right|^{2}+\frac{2 C_{1}}{h_{1}} , {\rm{故有}}\\ \;\;\;\sup \limits_{0 \leqslant s \leqslant t} \mathrm{e}^{\varepsilon s} E|\boldsymbol{x}(s)|^{2} \leqslant \frac{2 C_{1}}{h_{1}}+2 \mathrm{e}^{\varepsilon \tau} \kappa^{2}(1-\overline{\boldsymbol{\tau}})^{2} \sup \limits_{0 \leqslant s \leqslant t} \mathrm{e}^{\varepsilon s}\cdot \\ E|\boldsymbol{x}(s)|^{2}+2 \kappa^{2}(1-\bar{\tau})^{2} \mathrm{e}^{\varepsilon \tau} \sup \limits_{-\tau \leqslant s \leqslant 0}\|\phi\|^{2}\end{array}$
$ \kappa(1-\bar{\tau})<\sqrt{\frac{1}{2}}$,令ε4为1-2eε4τκ2(1-τ)2= 0的唯一解,则对任意0<εε1ε2ε3ε4,有
$\begin{array}{l} \;\;\;\sup \limits_{0 \leqslant s \leqslant t} \mathrm{e}^{\varepsilon s} E|\boldsymbol{x}(s)|^{2} \leqslant \frac{\frac{2 C_{1}}{h_{1}}+2 \kappa^{2}(1-\bar{\tau})^{2} \mathrm{e}^{\varepsilon \tau} \sup \limits_{-\tau \leqslant s \leqslant 0}\|\varphi\|^{2}}{1-2 \kappa^{2}(1-\bar{\tau})^{2} \mathrm{e}^{\varepsilon \tau}}:=\\C_{2} \end{array}$
即当t∈[0, ∞)时,eεtE|x(t)|2C2,即E|x(t)|2C2eεt
对于任意的${\bar q}$ ∈[2, ${\bar w}$),由Hölder不等式得到$ E|\boldsymbol{x}(t)|^{\bar{q}} \leqslant\left(E|\boldsymbol{x}(t)|^{2}\right)^{\frac{\bar{w}-\bar{q}}{\bar{w}-2}}\left(E|\boldsymbol{x}(t)|^{\bar{w}}\right)^{\frac{\bar{q}-2}{\bar{w}-2}} $
由引理1知$ C_{3}:=E|\boldsymbol{x}(t)|^{\bar{w}}<\infty$,故
$E|\boldsymbol{x}(t)|^{\bar{q}} \leqslant C_{3}^{\frac{{\bar{q}}-2}{{\bar{w}}-2}}\left(C_{2} \mathrm{e}^{-\varepsilon t}\right)^{\frac{\bar{w}-\bar{q}}{\bar{w}-2}} \leqslant C_{4} \mathrm{e}^{-\varepsilon t \frac{\bar{w}-\bar{q}}{\bar{w}-2}}$
所以式(17)成立。特别地,当${\bar q}$=2时,有式(18)成立。

3 例子

考虑一维HNSDDE
$\begin{array}{l} \;\;\;\mathrm{d}[x(t)-N(x(t-\tau(t)), t, r(t))]=f(x(t), \\x(t-\tau(t)), t, r(t)) \mathrm{d} t+g(x(t), x(t-\tau(t)), t, \\r(t)) \mathrm{d} B(t), t \geqslant 0 \end{array}$
其中
$\begin{aligned}& f(x, y, t, 1)=0.5 x+y^{3}-6 x^{3} \\& f(x, y, t, 2)=x+y^{3}-4 x^{3} \\& g(x, y, t, 1)=g(x, y, t, 2)=0.5 y^{2} \\& \tau(t)=0.1(1-\cos t), N(y)=0.1 y\end{aligned}$
显然f, g不满足线性增长条件。
r(t)为一个连续的马氏链,状态空间S={1, 2},算子$ \varGamma=\left(\begin{array}{cc}-2 & 2 \\ 1 & -1\end{array}\right), B(t)$为标准布朗运动且独立于r(t)。定义初值x(u)=0.2+cos u, u∈[-0.2, 0], r(0)=2。
由文献[10]可知系统(30)不稳定,以下将通过引入一个反馈控制函数使系统稳定。
增加控制函数u(x, t, 1)=-x, u(x, t, 2)=-2x,增加控制函数后系统(3)的具体形式为
$\begin{aligned}& \;\;\;\mathrm{d}[x(t)-0.1 x(t-\tau(t))]= \\& \left\{\begin{array}{c}\left(\frac{1}{2} x(t)+(x(t-\tau(t)))^3-6 x(t)^3-x(t-\right. \\\;\;\;\delta(t))) \mathrm{d} t+\frac{1}{2}(x(t-\tau(t)))^2 \mathrm{~d} B(t), i=1 \\\left(x(t)+(x(t-\tau(t)))^3-4 x(t)^3-2 x(t-\right. \\\;\;\;\delta(t))) \mathrm{d} t+\frac{1}{2}(x(t-\tau(t)))^2 \mathrm{~d} B(t), i=2\end{array}\right\} \\&\end{aligned}$
其中δ(t)=τ(t)。以下验证假设1~8。
假设1显然成立。令m1=3, m2=2,可知假设2成立。令λ4=2,可知假设8成立。假设3对如下常数成立:δ=τ=0.2, δ =τ =0.1,且假设4对$ \kappa=\frac{1}{9}$成立。
U1(x, t)=V(x, i, t)=|x|6, U2(x, t)=2.2x6+ x8,由Young不等式可得
$\begin{array}{l}\;\;\;\mathrm{L} V(x, y, t, i)+V_{x}(x-N(y), t, i) u(z, t, i) \leqslant\\\sup \limits_{x \in \mathbb{R}}\left(43 x^{6}-0.229 x^{8}\right)-8 \times U_{2}(x, t)+\frac{58}{9} \times(1-\bar{\tau}) \times\\U_{2}(y, t-\tau(t))+\frac{10}{9} \times(1-\bar{\delta}) \times U_{2}(z, t-\delta(t)) \end{array}$
故假设5对
$\begin{array}{l}\;\;\;\;c_{1}=\sup \limits_{x \in \mathbb{R}}\left(43 x^{6}-0.229 x^{8}\right)<\infty, c_{2}=8, c_{3}=\frac{58}{9}, \\c_{4}=\frac{10}{9}\end{array}$
成立。取p=4, q1=3, 可知假设6成立。取$ U(x, t, i) = \left\{ {\begin{array}{*{20}{l}} {2{x^2} + {x^4}, }&{i = 1}\\ {{x^2} + {x^4}, }&{i = 2} \end{array}} \right.$,再由Young不等式,令λ1=0.05, λ2=0.1, λ3=4可得
$\begin{array}{l} \;\;\;\ell U(x, y, t, i)+\lambda_{1}\left|U_{x}(\hat{\boldsymbol{x}}(t), t, i)\right|^{2}+\lambda_{2} \cdot\\|f(x, y, t, i)|^{2}+\lambda_{3}|g(x, y, t, i)|^{2} \leqslant-1.845|x|^{2}+\\0.369(1-\bar{\tau})|y|^{2}-6\left(x^{4}+x^{6}\right)+0.955 \times(1-\bar{\tau}) \times\\6\left(y^{4}+y^{6}\right)\end{array}$
若令H(x)=6(x4+x6), λ=1.845, β=0.369, α=0.955,则假设7成立。根据定理1条件发现κ, τ取值合理,进而可以得到δ≤0.057 6时,定理1所有条件成立,故对∀${\bar w}$ ≥6, ∀${\bar q}$ ∈[2, ${\bar w}$),存在ε>0使得
$\mathop {\lim }\limits_{t \to \infty } \sup \frac{1}{t} \ln \left(E|x(t)|^{\bar{q}}\right) \leqslant-\varepsilon \frac{\bar{w}-\bar{q}}{\bar{w}-2}$
特别地,${\bar q}$ =2时有$ \mathop {\lim }\limits_{t \to \infty } \sup \frac{1}{t} \ln \left(E|x(t)|^{2}\right) \leqslant-\varepsilon$

4 结论

本文采用函数方法,受文献[5]的启发在多项式增长的条件下讨论了变时滞反馈控制下的HNSDDEs的指数稳定性。最后,用一个例子证明了结论的有效性。
1
MAO X R , YUAN C G . Stochastic differential equations with Markovian switching[M]. London: Imperial College Press, 2006.

2
FEI W Y , HU L J , MAO X R , et al. Delay dependent stability of highly nonlinear hybrid stochastic systems[J]. Automatica, 2017, 82, 165- 170.

DOI

3
FEI C , SHEN M X , FEI W Y , et al. Stability of highly nonlinear hybrid stochastic integro-differential delay equations[J]. Nonlinear Analysis: Hybrid Systems, 2019, 31, 180- 199.

DOI

4
HU L J , MAO X R , SHEN Y . Stability and boundedness of nonlinear hybrid stochastic differential delay equations[J]. Systems & Control Letters, 2013, 62, 178- 187.

5
WU A Q , YOU S R , MAO W , et al. On exponential stability of hybrid neutral stochastic differential delay equations with different structures[J]. Nonlinear Analysis: Hybrid Systems, 2021, 39, 100971.

DOI

6
SHEN M X , FEI W Y , MAO X R , et al. Stability of highly nonlinear neutral stochastic differential delay equations[J]. Systems & Control Letters, 2018, 115, 1- 8.

7
SHEN M X , FEI C , FEI W Y , et al. Boundedness and stability of highly nonlinear hybrid neutral stochastic systems with multiple delays[J]. Science China Information Sciences, 2019, 62, 202205.

DOI

8
LI X Y , MAO X R . Stabilisation of highly nonlinear hybrid stochastic differential delay equations by delay feedback control[J]. Automatica, 2020, 112, 108657.

DOI

9
周之薇, 宋瑞丽. 变时滞反馈控制的混合中立型随机延迟微分方程的稳定性[J]. 井冈山大学学报(自然科学版), 2022, 43 (3): 6- 14.

ZHOU Z W , SONG R L . Stabilization of the hybrid neutral stochastic differential equations controlled by the time-varying delay feedback[J]. Journal of Jinggangshan University (Natural Science), 2022, 43 (3): 6- 14.

10
SHEN M X , FEI C , FEI W Y , et al. Stabilisation by delay feedback control for highly nonlinear neutral stochastic differential equations[J]. Systems & Control Letters, 2020, 137, 104645.

11
CHEN W M , XU S Y , ZOU Y . Stabilization of hybrid neutral stochastic differential delay equations by delay feedback control[J]. Systems & Control Letters, 2016, 88, 1- 13.

文章导航

/